AivexaNewsSearch
AI news for builders and product teamsChecked every hour

Zest’s new AI agent finds restaurants and books reservations over text

Collected Oct 30, 2026

Zest, a restaurant discovery app that launched earlier this year, introduced a personal AI restaurant concierge on Thursday. The feature lives inside text messages rather than a standalone app, and it can both recommend places and complete the booking. Users reach it by texting (212) 470-6067. The company was founded in November 2024 and had raised $1.8 million in pre-seed funding from Alexis Ohanian at 776 and Steve Jang at Kindred Ventures.

The core idea is to collapse discovery and reservation into one conversation. A user texts the concierge, which responds with suggestions drawn from their dining history, then handles checking availability and booking a table through connected services: OpenTable, Resy and SevenRooms. The feature was tested internally last month but held back until booking capability was ready, according to the company.

What makes Zest's recommendations different from a general-purpose assistant is where the data comes from. The app links to users' credit cards through Plaid, the financial data provider used by banks and budgeting apps, to see where they actually eat. Zest says it looks only at food and drink transactions when building personalized maps and lists. Founder and CEO Mario Gomez-Hall framed this as a speed advantage: the agent answers from a user's own history and their friends', rather than needing weeks of conversation to learn preferences. In practice, this is a cold-start shortcut. Most recommendation systems have to infer taste from explicit ratings or a long history of in-app behavior; payment data provides a ready-made record of real visits, which is a stronger signal for restaurant choice than clicks. The trade-off is dependency on a sensitive data category and on a third-party aggregator, and a likely ceiling for users who are uncomfortable connecting cards.

Group decision-making gets specific treatment. Zest can be added to a group chat, where it blends the tastes of everyone present, runs an in-thread vote to settle disagreement, and then books a table without anyone leaving the conversation. That voting step is a small but meaningful design choice: it keeps the agent as a facilitator rather than an autonomous decision-maker, which matters when the reservation is for several people with different preferences.

Recommendations can be filtered by cuisine, craving, occasion, dietary restrictions or neighborhood. The agent also answers practical questions about a restaurant, such as how busy it might be or whether it offers Wi-Fi, and can send the menu to help someone decide.

In TechCrunch's testing, the agent performed reasonably well, returning suggestions matched to taste and to conditions like time of day, party size and location. The publication noted that table-finding has existed in apps for years, but that the back-and-forth over options with immediate availability times is where the convenience sits. That is a fair read: the underlying inventory comes from established booking networks, so the novelty is the interface and the personalization layer on top, not new supply.

On the business side, Gomez-Hall described booking and ordering as the first revenue lines, with food ordering planned. Zest currently books through OpenTable or Resy, and the company sees transaction participation as the monetization path. He argued the taste data compounds: because the agent starts from where a user has actually been rather than from a questionnaire, it learns faster than a general assistant, which he said is valuable beyond dinner. The company also hopes the shift to texting helps it grow past its initial thousands of users.

Longer term, Zest is considering extending the concierge beyond restaurants to recommendations for businesses, shopping and more. That ambition is early-stage; the announced capability today is restaurants, with booking through the three named platforms.

Why it matters: This is another signal that consumer AI products are migrating from app icons to messaging threads, where the agent handles a task end to end instead of handing users off to a separate booking flow. For developers building in commerce and local services, the interesting parts are the pattern of importing existing behavioral data to skip onboarding, group voting as a way to distribute decisions, and acting as a thin layer over incumbent reservation APIs rather than replacing them. Teams weighing a similar build should note, as inference rather than stated fact, that credit-card-linked personalization carries privacy and compliance overhead, and that dependence on OpenTable, Resy and SevenRooms inventory limits how differentiated the actual booking experience can be.

Read at TechCrunch · AI

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

Can't agree on where to eat? Zest's AI concierge can combine everyone's tastes, put restaurant options to a vote, and make the reservation.